AI Summary
5 min readWhen an AI System Improves Itself—Not Just Automates Its Own Development
The term "recursive self-improvement" (RSI) gets thrown around a lot in AI, but most uses of it are wrong. As a16z general partner Martin Casado learns in conversation with Maker Maker AI co-founders Daval Ajada and Owen He, the real thing is far rarer and harder to demonstrate than the hype suggests. The key distinction: are you using AI to help build better AI (which humans have done for decades with every technology), or is the system being improved also the system doing the improving, in a closed loop?
What RSI Actually Means—and What It Doesn't
The founders are blunt about the definitional confusion. "Many people when they say RSI, they mean different things," Daval explains. The critical question is: "What is the optimizer and what is the optimizee?" When a human uses AI tools to develop better AI, that's not RSI—it's autocatalysis, the same phenomenon that has driven technological progress since the steam engine (you need coal to mine coal faster). The improver is still a human brain.
Continue reading the full summary in the app — free to try.
Read Full Summary →Free • No credit card required
Never miss an episode of a16z Show
Get every new episode summarized in your inbox — free, ~5 minutes to read.
No spam. Unsubscribe anytime.
What you'll learn
- 1 (00:00) **Introduction and Defining RSI** - The episode opens with the core idea: building AI that improves itself, and host Martin Casado sets up the conversation with Maker Maker co-founders Daval Ajada and Owen He.
- 2 (02:03) **Guest Introductions and Backgrounds** - Daval and Owen introduce themselves, their academic backgrounds in physics and math, and their path to working on RSI.
- 3 (03:17) **What RSI Is (and Isn’t)** - The hosts distinguish true RSI from the common practice of using AI to automate pieces of AI development.
- 4 (05:44) **The Closed Loop and the Math of Improvement** - The hosts explain the fundamental difference between autocatalysis and a closed-loop system, using a simple mathematical model.
- 5 (07:38) **From Closed Loop to Takeoff: An Empirical Question** - The discussion moves from the theoretical potential of RSI to the practical reality that its behavior is an empirical, not analytical, problem.
- 6 (09:26) **The Hybrid System: Humans as Agents** - The hosts clarify that their approach to RSI includes humans as a special kind of agent within a multi-agent system.
- 7 (10:48) **Measuring Real Progress: The Denominator Problem** - The conversation focuses on how to avoid fooling yourself by conflating RSI with simply throwing more resources at a problem.
+ Full timestamped outline available in the app
Guests on this episode
Show Notes
a16z General Partner Martin Casado sits down with MakerMaker.AI co-founders Dhaval Adjodah and Owen He to unpack recursive self-improvement, or RSI, and separate the increasingly popular term from what it would actually mean for an AI system to improve itself.
Dhaval and Owen distinguish true RSI from simply using AI to automate pieces of AI development. Their goal is a closed loop where the system being improved is also doing the improving, across components including models, harnesses, and eventually more of the AI stack. They explain why closing that loop still doesn’t guarantee runaway “takeoff,” and why measuring real progress requires accounting for both capability gains and the resources required to achieve them.
They also share what they’re seeing in their own research, including early signs that self-improving systems can increase performance while improving efficiency. And they dig into the practical challenges of proving those gains are real, from variance and reward hacking to the enormous difficulty of distinguishing genuine improvement from noise.
Resources:
Follow Dhaval Adjodah on LinkedIn: https://www.linkedin.com/in/rand0m
Follow Owen He on LinkedIn:https://www.linkedin.com/in/owen-he-b7b064279/
Follow Martin Casado on X: https://x.com/martin_casado
Stay Updated:
Find a16z on YouTube: YouTube
Find a16z on X
Find a16z on LinkedIn
Listen to the a16z Show on Spotify
Listen to the a16z Show on Apple Podcasts
Follow our host: https://twitter.com/eriktorenberg
Please note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see a16z.com/disclosures.
Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.